Google Brings AI Music Generation to Gemini—We Tried It, and It's Too Little Too Late
Google DeepMind’s Lyria 3, an AI music generation model, is now available in beta to Gemini app users aged 18 and over. Users can generate fully produced, 30-second tracks—complete with lyrics, instrumentals, and AI-generated cover art—by describing an idea or uploading a photo. While Lyria 3 creates coherent, catchy samples for mainstream genres like pop, R&B, and hip-hop, it struggles with unusual or specific styles and cannot generate full-length songs like competitors Suno or Udio. Both Suno and Udio offer greater customization, longer track lengths, and more advanced features, but have faced copyright lawsuits related to training data use. Google claims Lyria 3 was developed with close attention to copyright, intentionally avoids imitating specific artists, and embeds every track with SynthID AI watermarks for transparency. Lyria 3 is available on desktop and will soon be accessible on mobile, with higher generation limits for paid Gemini tiers. It is also expanding YouTube’s Dream Track AI soundtrack feature to creators worldwide.
